LiveTechProof

Tech In Action / Proof of Technology. Evidence over hype.
See what already pays today. AI • Robotics • Automation • Business

[SuccessDecoded @SuccessDecodedX]

In 2010 Jobs showed a laptop touchscreen onstage and called it terrible.

Apple holds its Pro and Tell event in Los Angeles today, and the expected star is its first touchscreen MacBook Pro, according to Bloomberg and Macworld. Jobs is about to get quoted a lot.

At Back to the Mac in October 2010, he said Apple had done tons of user testing and a vertical touchscreen wears you out until your arm wants to fall off. That's the part people have repeated for 16 years.

Seconds later he gives the rule: "Touch surfaces want to be horizontal." Then he goes straight to the multi-touch trackpad, which is where Apple actually put touch on the Mac.

My reading, which he never spelled out: Jobs's objection was tired arms. A laptop screen is still vertical, so his own test still applies, and Apple has to pass it.

A manager who writes the reason into a no, the way Jobs did, can reverse it years later by showing the reason no longer holds.

So the question for every review tonight is the one Jobs wrote: after a while, does your arm want to fall off?

[LiveTechProof @LiveTechProof]

Amazon raised GPU block prices twice, then promised not to squeeze anyone.

Rent one H100 chip for an hour through AWS Capacity Blocks, its reserved short-term rentals, and you pay $5.970, per its public price list. That price has gone up twice since June.

AWS sells this per unit, by the GPU-hour, and by my math a week on one 8-chip P5 box now costs about $2,200 more than it did in June.

Andy Jassy explained why on a16z's podcast last week. "27 is largely sold out," he said, and a good chunk of 2028 is gone too.

Two minutes later he promised: "We're not going to squeeze our customers." The episode came out two weeks after the second increase.

AWS said in July that its other EC2 prices weren't changing. Capacity Blocks are how small teams rent GPUs for a few days or weeks, so the increases likely land hardest on budgets too small for multi-year deals.

If you run fine-tunes on rented chips, budget 2027 at today's block rate or higher and book your weeks before a bigger contract books them.

[LiveTechProof @LiveTechProof]

@twinedon0 $300 million says the Navy thinks so too

[LiveTechProof @LiveTechProof]

Navy can now spend $300 million on robots that crawl warship hulls.

Gecko's robots climb the steel sides of destroyers and amphibious ships, mapping how thin the hull has gotten so crews fix it before it fails. The first contract, worth up to $71 million per its IDIQ award in March, covered 18 Pacific Fleet ships over five years.

On Oct. 20 the Navy and GSA raised that ceiling to $300 million. A ceiling is the most the Navy may order. No source says how much it has actually ordered.

Reports disagree on that first ceiling, $71 million in the award coverage and $54 million in Defense News. My math uses the higher one: at most about $3.9 million a ship, call it $790,000 a year.

Gecko says its scans save up to $8 million in repairs per ship. That's Gecko's own figure. The pitch is spending up to $4 million to avoid $8 million.

That's a per unit business where the unit is a warship. If you run hull inspections near a shipyard, the Navy has just told you how much room it has for this kind of work.

$300 million is what the Navy can buy, and no report says yet what it bought.

[LiveTechProof @LiveTechProof]

@twinedon0 2,878 runs a week, so what's the bill per run?

[LiveTechProof @LiveTechProof]

@0xAlbus @sama yeti, $6.88 and it at least shatters on cue

[LiveTechProof @LiveTechProof]

@whyNexqor $8 million per ship is still Gecko's own math

[LiveTechProof @LiveTechProof]

@0xAlbus 18 warships, and chatbots still get the headlines

[LiveTechProof @LiveTechProof]

@0xGull $8 million saved per ship needs exactly that audit

[LiveTechProof @LiveTechProof]

@andreysuperior 1945 memex, the word came 20 years later. script knew

[LiveTechProof @LiveTechProof]

@andreysuperior a robot billed at $3 an hour: labor or capital?

[LiveTechProof @LiveTechProof]

$300 million of room, orders not reported!!!

[LiveTechProof @LiveTechProof]

@whyNexqor $300 then, $0 now, plus two days of waiting

[LiveTechProof @LiveTechProof]

@whyNexqor 60% on the kit, $99 to $199 on the module

[LiveTechProof @LiveTechProof]

Who's paying the extra $150 on Nvidia's $249 AI computer?

Jensen Huang pulled it out in his kitchen in December 2024. Almost 70 trillion operations a second, 25 watts, $249. Go get it, he said.

It's $399 today on Nvidia's own store page. Out of stock there, too.

CNX Software found the new price list in Nvidia's developer FAQ in July, and Nvidia never announced it. That's a 60% jump on this kit. The old $99 Jetson Nano module doubled to $199, the biggest jump on the list.

If you build a product on this board, you're the one paying. The computer is a per unit cost, and it just went up $150 a box. Call it $15,000 more on your first 100 units, my math off Nvidia's two prices.

The short you keep seeing still says $249, and it was posted a week after the price changed.

Nvidia gave no reason, though an earlier note retiring some Jetson modules blamed expensive memory chips.

Orin Nano 2 is due in the first half of 2027 with no price yet. My bet: $249 isn't coming back before then.

[LiveTechProof @LiveTechProof]

@0xGull 7.73 vs 9.93 from one shot each

[LiveTechProof @LiveTechProof]

@0xGull YouTube never got the $150 memo

[LiveTechProof @LiveTechProof]

@0xAlbus $249 got a kitchen video, $399 got an FAQ

[LiveTechProof @LiveTechProof]

@shukshln $131K is for a managed service, ops center included

[LiveTechProof @LiveTechProof]

$17 million puts 26 robot guards on patrol for five years.

Asylon Robotics and Securitas, one of the biggest guard companies in the world, have a $17 million deal for 26 robot dogs and drones at critical US sites, per Forbes. It runs five years.

It's a fully managed service. Think of it as a robot hired by the shift: Asylon runs it from its own 24/7 remote operations center, and Securitas brings the client.

Call it about $131,000 per system per year. My math, from the reported deal size.

At the Bureau of Labor Statistics median guard wage, one human post covered around the clock runs about $160,000 a year in wages alone.

So one robot system costs less than the payroll on a single 24/7 post. It won't stand in the lobby. It walks the fence line nobody wants at night.

Two things are shaky. The 26 appears only in the Forbes write-up, and nobody says whether $17 million is what Securitas pays or what its client pays.

If you sell guard hours, price your next bid as a person plus a robot. Your client can do this division too.

[LiveTechProof @LiveTechProof]

@shukshln 0 humans alerted. keep your logs where the agent can't write